Predicting Programmers' Ability to Modify Software.
Abstract
This report describes the second experiment in a program of research designed to identify characteristics of computer software which are related to its psychological complexity. Thirty-six experienced programmers were given unlimited time to make specified modifications to a preliminary program and three experimental programs. The correctness of the modification and the time required to make each modification served as dependent variables. Results indicated that the difficulty of the modification was significantly related to the time to solution. This relationship was described by a hyperbolic function relating time to the number of statements to be inserted in the code. Modest effects on the score and time were observed for order presentation, suggesting a learning effect. On two of the three programs studied, better modifications were made when the control flow of the original programs was well-structured. No performance effects were related to the absence of comments or the type of comments (in-line versus global) used. Moderate relationships with the criteria were observed for several complexity metrics, and these were strongest where metrics were obtained from the modified code rather than the original programs.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1978
- Accession Number
- ADA056079
Entities
People
- B. Curtis
- L. T. Love
- M. A. Borst
- S. B. Sheppard
Organizations
- General Electric